Pulmonary Vasodilator and Inodilator Drugs in Cardiac Surgery: A Systematic Review With Bayesian Network Meta-Analysis

Abstract

Objective: The authors performed a systematic review to evaluate the effect of pharmacologic therapy on pulmonary hypertension in the perioperative setting of elective cardiac surgery (PROSPERO CRD42023321041). Design: Systematic review of randomized controlled trials with a Bayesian network meta-analysis. Setting: The authors searched biomedical databases for randomized controlled trials on the perioperative use of inodilators and pulmonary vasodilators in adult cardiac surgery, with in-hospital mortality as the primary outcome and duration of ventilation, length of stay in the intensive care unit, stage 3 acute kidney injury, cardiogenic shock requiring mechanical support, and change in mean pulmonary artery pressure as secondary outcomes. Participants: Twenty-eight studies randomizing 1,879 patients were included. Interventions: Catecholamines and noncatecholamine inodilators, arterial pulmonary vasodilators, vasodilators, or their combination were considered eligible interventions compared with placebo or standard care. Measurements and main results: Ten studies reported in-hospital mortality and assigned 855 patients to 12 interventions. Only inhaled prostacyclin use was supported by a statistically discernible improvement in mortality, with a number-needed-to-treat estimate of at least 3.3, but a wide credible interval (relative risk 1.26 × 10-17 - 0.7). Inhaled prostacyclin and nitric oxide were associated with a reduction in intensive care unit stay, and none of the included interventions reached a statistically evident difference compared to usual care or placebo in the other secondary clinical outcomes. Conclusions: Inhaled prostacyclin was the only pharmacologic intervention whose use is supported by a statistically discernible improvement in mortality in the perioperative cardiac surgery setting as treatment of pulmonary hypertension. However, available evidence has significant limitations, mainly the low number of events and imprecision.